Advancements in Synthetic Video Generation for Autonomous Driving
• Ex is the expected value of real input .
• Ez is the expected value of generated fake instance .
• D ( x ) is the expected estimate of the discriminator for the real instance to be fake .
• D ( G ( z )) is the expected estimate of the discriminator for the fake instance to be real .
4.4 GENERATOR ARCHITECTURE
The generator consists of a Multi SEAN-based image generator in which inputs are segmented as Image embedding , Flow and Level embedding as well previous frames outputs . Figure 4-1 and Figure 4-2 give an overview of our generator architecture . The generator comprises a series of Multi SEAN residual blocks and up-sampling layers . The structure of each Multi SEAN residual block is shown in Figure 4-2 , which replaces the SEAN layers in the original SEAN residual blocks with Multi SEAN layers .
Figure 4-1 : Multi SEAN-based image generator .
Journal of Innovation 119